Cosmic FL cuts onboard satellite learning costs by pruning experts
Resource-Aware Federated Mixture-of-Experts with Adaptive Pruning for Onboard Learning in LEO Satellite Constellations
Machine LearningComputer Vision and Pattern Recognition
Summary
Satellites in low Earth orbit need to learn from the data they collect, but they have limited power, memory, and communication ability. The authors propose COSMIC-FL, a new way to split learning tasks among small, specialized parts of a model, so each satellite trains only what matters to its own data. This method also shrinks the models intelligently to save computing and communication resources. Experiments show that COSMIC-FL keeps learning accurate while greatly reducing energy use and data sent between satellites and ground stations.
What this means in practice
- •For satellite systems engineers: Develop onboard learning setups that run efficiently on satellites with varied data and limited resources, reducing communication with ground stations.
- •For embedded device developers: Optimize machine learning models for resource-constrained devices by combining expert routing and adaptive model pruning to save energy and computation.
Authors
Mohamed Shaaban, Mohamed Elmahallawy, Marius Bernahrndt, Tobias Hecking
Abstract
Low-Earth-orbit (LEO) satellites are increasingly expected to perform onboard learning for applications such as disaster response and environmental monitoring. However, conventional federated learning (FL) is ill-suited to onboard satellite learning, as it assumes computational, memory, and communication resources beyond the capabilities of resource-constrained LEO platforms, often necessitating the transmission of raw imagery to ground stations. We present COSMIC-FL, a resource-aware FL framework for efficient onboard learning in LEO satellite constellations. COSMIC-FL introduces two complementary Mixture-of-Experts (MoE) architectures: a Sliced design that shares backbone representations while activating task-specific channel subsets, and a Modular design that employs lightweight gating to route inputs to physically separated expert networks. A semantic class-to-expert mapping enables each satellite to train, update, and communicate only the expert paths relevant to its local data. To further improve efficiency, COSMIC-FL integrates staged optimization with three structured pruning strategies: server-side pruning, client-side fixed-ratio pruning with mean-vote aggregation, and adaptive client-side per-layer pruning based on aggregated importance and a MAD-based gap criterion. Combined with semantic expert routing, these techniques jointly adapt computation and model sparsity to both data semantics and layer importance, yielding a favourable accuracy--efficiency trade-off for heterogeneous space platforms. Experiments on six image classification benchmarks under highly non-i.i.d. settings show that COSMIC-FL maintains competitive accuracy while reducing communication, computation, and energy consumption by up to 80% over SOTA FL methods. We further validate COSMIC-FL on an NVIDIA Jetson AGX Orin, confirming its efficiency gains under realistic embedded deployment constraints.